Nicholas’ Status Report for 8th February 2025
I have been researching which ML Model we should use for our project, as well as confirming my belief that there are no alternate Computer Vision techniques we can use. In my research, I reviewed edge detection techniques and color analysis systems for detecting cards. However, as I initially believed, these systems are not robust to different conditions, which is a user requirement for us. Furthermore, these systems do not do further feature processing and refinement like the ML Model’s we will be using can do. This lack of further processing means that a pure Computer Vision approach would not meet the user requirements. Once I had fully confirmed my suspicions, I began to further research Resnet-50 and the YOLO family of models. We have currently decided to use a YOLO model, since after further research, we realized Resnet is a common backbone for most object detection pipelines, but would require further scaffolding on our part. YOLO would streamline this process, so for next week, we are deciding which YOLO model to use, and we will begin preparing the model and dataset for training.